Quantifying mRNA in postmortem human brain: influence of gender, age at death, postmortem interval, brain pH, agonal state and inter-lobe mRNA variance

Abstract

The quantification of mRNA in postmortem human brain is often made complicated by confounding factors. To assess the importance of potential confounders TaqMan™ real-time RT-PCR was used to measure seven mRNAs (β-actin, glyceraldehyde-3-phosphate dehydrogenase (GAPDH), cyclophilin, microtubule-associated protein (MAP) 2, neuron-specific enolase (NSE), glial fibrillary acidic protein (GFAP), amyloid precursor protein (APP) isoform 770) in cortical samples taken from 90 Alzheimer's disease (AD) and 81 control brains. Demographic data for the brain samples were assessed for interaction between factors and amounts of mRNA. Gender was found to play a role in that females had lower levels of mRNA relative to males; this was consistent in both the AD and control brains. Age at death had inconsistent but significant correlations to amounts of mRNA; male and female controls both had negative correlations, female AD a positive correlation and male AD no correlation. Positive correlations were found between brain pH and amount of mRNA in all genes except glial fibrillary acidic protein (GFAP); correlations were consistent across all groupings of pathology and gender. Mean brain pH was significantly lower in AD (6.4) than in control subjects (6.5, ANOVA, p<0.01), though there was no difference between male and females of either group. No correlation was found between brain pH and age at death. Postmortem interval was correlated with brain pH in Alzheimer's disease brains but not controls. Agonal state was generally a poor predictor of mRNA levels whilst inter-lobe variance of mRNA was found to be non-significant in control brains. Given that gender, age at death and brain pH all have significant effects upon mRNA levels it is recommended that these factors be taken into account when quantifying gene expression in postmortem human brain.

Introduction

The measurement of gene expression is important because it may indicate the dysregulation of genes implicated in pathogenesis. Quantifying a single mRNA in a postmortem human brain, however, is often made complicated by confounding factors. A confounder causes bias due to the variable of interest, in this case an mRNA level, also being related in some way to the confounding variable. For example, one study found a decrease of muscarinic receptor M₁ mRNA correlated to increasing antemortem duration of coma that was independent of diagnosis [9]. That is, both the controls and the Alzheimer's disease (AD) brains had the correlation. If an AD sample is then found to have more brains with long antemortem comas, then it is possible to assign change in the test variable to pathology rather than its correlate, the duration of coma.

Confounding is a frequent problem with biochemical measurements of postmortem human brain tissue. Among the many factors suggested as contributing to bias are: gender, age of death, postmortem interval, brain pH, antemortem medication, terminal coma, hypoxia, pyrexia, death-to-refrigeration interval, freezer storage time and handling of the body [2], [8], [9], [10], [11], [15], [19], [20]. As a result, it is common practice to use a matched design either by balancing groups for equal numbers of case and controls with potential confounding factors (stratum matching), or by deliberately match-pairing case and controls with the same values for confounders. Confounding can also be corrected for in the statistical analysis. The important point is to know what the confounders are and which ones are more influential than others.

In previous studies, brain tissue has been matched for age at death and postmortem delay [5], [20]. Further studies indicated that short postmortem interval is not the best or only correlate of mRNA preservation [2], and tissue pH may critically affect mRNA stability and recovery [14], [15]. Brain tissue from subjects dying after a terminal hypoxic episode has typically a lower pH than that where death is more acute [8], [9], [11], [14]. Recent studies have therefore prioritised pH-matching placing less emphasis on autopsy delay [18]. A review of over 20 studies of the effects of postmortem delay on specific mRNA levels showed a minimal effect with only one or two exceptions [2], but Barton et al. also stressed the need to be aware of agonal state and in particular hypoxia and pyrexia. The Barton et al. review identified a number of confounders of selective mRNA increases or decreases due to factors other than the disease under investigation, and caution in interpretation of postmortem data was indicated.

Factors to be considered in quantifying mRNA and in particular the normalisation of mRNA to housekeeper genes have been reported previously [3], [4]. In this study, we examined the role of gender, age at death, postmortem interval, brain pH and inter-lobe variance of mRNA and inter-variable relationships.